In general relativity, gravitation is described not as a Newtonian force acting instantaneously across space, but as the geometric curvature of four-dimensional spacetime caused by the presence of mass and energy.⁹ When massive objects accelerate, the local spacetime metric changes, and these variations propagate outward as transverse waves at the speed of light, denoted $c$.¹⁰ Unlike electromagnetic radiation, which is typically dominated by dipole emission arising from the oscillation of positive and negative electric charges, gravitational radiation cannot have a dipole component. Because negative mass does not exist and the conservation of linear and angular momentum forbids time-varying mass monopole and dipole moments in an isolated system, the lowest-order radiation allowed in general relativity is quadrupolar.

A system generates gravitational waves only when the second time derivative of its mass quadrupole moment is non-zero, meaning its mass distribution must accelerate in a non-spherically symmetric manner. A perfectly spherical star pulsating radially, or a perfectly axisymmetric disk spinning smoothly around its symmetry axis, emits no gravitational radiation. By contrast, two compact objects orbiting their common center of mass possess a continuously changing quadrupole moment, radiating gravitational waves at twice their orbital frequency in a circular orbit, or across multiple harmonics if the orbit is eccentric.¹¹ According to Einstein's quadrupole formula, the luminosity of this radiation scales with the fifth power of the inverse speed of light ($c^{-5}$) and the gravitational constant $G$, making gravitational-wave emission exceedingly weak for all but the most massive and relativistic systems.

As a gravitational wave propagates through flat spacetime, its physical effect in the transverse-traceless gauge is to rhythmically stretch and compress the proper distances between freely falling test particles perpendicular to the wave's direction of travel.¹²¹³ General relativity predicts exactly two independent linear polarization states for these tensor waves, designated "plus" ($+$) and "cross" ($\times$), where the axes of the cross polarization are rotated by 45 degrees relative to the plus polarization.¹⁴ If a plus-polarized wave passes perpendicularly through a circular ring of test particles, it deforms the ring into an ellipse elongated vertically and compressed horizontally during the first half-cycle, before reversing the deformation in the second half-cycle; a cross-polarized wave produces the same quadrupolar deformation along diagonal axes. Alternative metric theories of gravity permit up to four additional polarization modes—two vector modes and two scalar modes, known as breathing and longitudinal states—whose absence in observational data serves as a stringent test of general relativity.

The amplitude of a gravitational wave is quantified by the dimensionless strain $h$, defined as the fractional change in length ($\Delta L / L$) induced across a baseline of length $L$. Because strain decreases inversely with the distance $r$ from the source ($h \propto 1/r$), rather than with the inverse square of the distance as flux does, doubling the sensitivity of a gravitational-wave detector doubles the distance to which it can detect sources and increases the observable volume of the universe by a factor of eight. Even for cataclysmic astrophysical collisions, the strain reaching Earth typically measures $10^{-21}$ or smaller, corresponding to a displacement smaller than the diameter of an atomic nucleus across a baseline of several kilometers.¹⁵¹⁶¹⁷

The concept that changes in a gravitational field propagate at a finite speed rather than instantaneously was explored in the late nineteenth century by William Kingdon Clifford in 1876 and Oliver Heaviside in 1893, the latter examining analogies between gravitation and Maxwell's equations of electromagnetism. In June 1905, the French mathematician Henri Poincaré published a paper on relativistic electron dynamics in which he explicitly proposed that Lorentz invariance required gravitation to propagate at the speed of light through what he termed ondes gravifiques (gravitational waves).¹⁸ Following the completion of his field equations of general relativity in November 1915, Albert Einstein initially doubted in a February 1916 letter to Karl Schwarzschild that such waves existed, noting the absence of gravitational dipoles. By June 1916, however, Einstein linearized the field equations under a weak-field approximation and showed that they admitted wave solutions, correcting an error in his initial coordinate choice in a foundational 1918 paper that established the quadrupole formula.

For the next four decades, physicists debated whether gravitational waves possessed true physical reality and carried energy, or whether they were merely coordinate illusions ("gauge modes") that could be eliminated by mathematical transformations.¹⁹ In 1922, Arthur Eddington demonstrated that two of the three wave types Einstein had identified in 1916 could travel at arbitrary speeds depending on the coordinate system—prompting his remark that they propagated "at the speed of thought"—while confirming that transverse-traceless waves traveled strictly at the speed of light. In 1936, Einstein and his assistant Nathan Rosen submitted a manuscript titled "Do Gravitational Waves Exist?" to the journal Physical Review, concluding that exact plane-wave solutions to the full non-linear field equations could not exist without singularities. After an anonymous peer reviewer—later identified as the cosmologist Howard Percy Robertson—pointed out that the singularities were merely coordinate artifacts of cylindrical symmetry, Einstein, assisted by Leopold Infeld, revised the paper prior to its 1937 publication in the Journal of the Franklin Institute to describe cylindrical gravitational waves.

The theoretical controversy was largely resolved at the Chapel Hill conference (designated GR1) in North Carolina in January 1957.²⁰ At the meeting, Felix Pirani presented a covariant, coordinate-independent formulation of gravitational radiation using the Riemann curvature tensor to show how a passing wave induces measurable relative accelerations between nearby test particles. Drawing on Pirani's formalism, Richard Feynman—alongside Hermann Bondi—introduced the "sticky bead argument," a thought experiment imagining a rigid rod on which two beads are free to slide with slight friction.²¹ As a gravitational wave passes perpendicularly to the rod, the proper distance between the beads changes while the rod resists deformation through electromagnetic interatomic forces, causing the beads to rub against the rod and generate heat. This demonstrated conclusively that gravitational waves transfer energy to matter and must be physically detectable.

Inspired by the Chapel Hill discussions, the American physicist Joseph Weber at the University of Maryland designed and built the first experimental gravitational-wave antennas in the 1960s. Known as Weber bars, these detectors consisted of solid aluminum cylinders measuring roughly 1.5 meters in length and 66 centimeters in diameter, weighing about three metric tons, and fitted with piezoelectric sensors to detect resonant mechanical vibrations around 1,660 hertz. In 1969, Weber reported coincident signals between bars located in Maryland and at Argonne National Laboratory near Chicago, claiming the detection of gravitational radiation from the Galactic Center. Throughout the 1970s, independent research groups in the United States, Europe, the Soviet Union, and Japan constructed more sensitive, cryogenically cooled resonant bars, none of which corroborated Weber's claims, leading to a scientific consensus that his signals were instrumental noise.

Conclusive indirect proof of gravitational waves came from radio astronomy.²² In 1974, Russell Alan Hulse and Joseph Hooton Taylor Jr., using the 305-meter Arecibo Observatory in Puerto Rico, discovered PSR B1913+16, the first known binary pulsar, located approximately 21,000 light-years from Earth.²³ The system consists of an active radio pulsar with a pulse period of 59 milliseconds locked in a tight 7.75-hour elliptical orbit with a companion neutron star. By timing the arrival of the radio pulses over several years, Taylor and Joel M. Weisberg demonstrated that the system's semimajor axis shrinks by 3.5 meters per year and its orbital period decreases by 76.5 microseconds per year, matching general relativity's prediction for energy loss via gravitational radiation to within a fraction of a percent. Hulse and Taylor were awarded the 1993 Nobel Prize in Physics for the discovery.

Like electromagnetic radiation, gravitational waves span a broad spectrum of frequencies determined by the masses and dynamical timescales of their sources, ranging from roughly $10^{-16}$ hertz to several kilohertz.²⁴²⁵ Because no single detector can cover this entire span, astrophysicists classify gravitational waves into four primary categories based on the physical mechanisms that generate them and the morphology of their waveforms: compact binary coalescences, continuous waves, burst signals, and the stochastic background.²⁶

Compact binary coalescences (CBCs) are the most prominent and frequently observed sources, arising from pairs of dense stellar remnants—black holes, neutron stars, or white dwarfs—locked in decaying orbits. The evolution of a coalescing binary proceeds through three distinct phases. During the prolonged inspiral phase, the two bodies orbit at wide separations, losing orbital energy and angular momentum to gravitational radiation; as they draw closer, their orbital velocity and the amplitude and frequency of the emitted waves steadily increase, producing a characteristic "chirp" waveform. When the orbits shrink to the innermost stable circular orbit, the system enters the highly non-linear merger phase, where the objects plunge together at speeds approaching a significant fraction of the speed of light, generating peak gravitational-wave luminosity. Finally, the newly formed remnant—typically a deformed black hole—settles into a stationary Kerr state during the ringdown phase by emitting damped sinusoidal waves known as quasi-normal modes, whose frequencies and damping times depend strictly on the remnant's final mass and spin. While stellar-mass binaries (1 to 100 solar masses) merge in the high-frequency band of 10 to $10^{4}$ hertz, supermassive black hole binaries ($10^{5}$ to $10^{10}$ solar masses) radiate in the millihertz and nanohertz regimes.²⁷

Continuous gravitational waves are quasi-monochromatic signals of nearly constant frequency and amplitude emitted over long durations. The primary astrophysical candidates for continuous emission in the audio frequency band are rapidly spinning neutron stars (pulsars) that possess slight non-axisymmetric deformations, such as crust "mountains" a few millimeters high supported by elastic stresses or intense internal magnetic fields. In the millihertz band, tens of thousands of detached white dwarf binaries within the Milky Way also act as continuous sources long before they approach merger.

Burst gravitational waves are short-duration, transient signals whose waveforms are either poorly modeled or result from complex, chaotic physics. Expected sources of gravitational-wave bursts include the asymmetric core collapse of massive stars during Type II and Type Ib/c supernovae, magnetar flares and pulsar starquakes ("glitches"), and the cosmic strings or hyperbolic encounters of compact objects in dense star clusters.²⁸²⁹ For a core-collapse supernova to emit detectable gravitational waves, the collapse must deviate from spherical symmetry through rotational flattening, convection, or standing accretion shock instabilities.

The stochastic gravitational-wave background is a persistent, random superposition of unresolved waves arriving from all directions in the sky, analogous to the cosmic microwave background. It comprises both an astrophysical foreground—formed by the combined emissions of countless distant compact binary mergers and supermassive black hole pairs across cosmic history—and a primordial cosmological background. Primordial gravitational waves are theorized to have originated in the first fractions of a second after the Big Bang through quantum fluctuations amplified during cosmic inflation, first-order cosmological phase transitions, or topological defects in the early universe.³⁰

On September 14, 2015, during the preparations for its first observing run (O1), the newly upgraded Advanced LIGO detectors recorded the event designated GW150914, announced publicly on February 11, 2016.³⁶ The signal, which swept upward in frequency from 35 to 250 hertz over 0.2 seconds with a peak strain of $1.0 \times 10^{-21}$, originated from the inspiral and merger of two black holes of approximately 36 and 29 solar masses at a luminosity distance of 1.3 billion light-years. The collision formed a single spinning black hole of 62 solar masses, radiating the energy equivalent of 3 solar masses ($5.3 \times 10^{47}$ joules) in gravitational waves within a fraction of a second—briefly exceeding the combined electromagnetic luminosity of all stars in the observable universe. The observation provided the first direct proof of the existence of binary stellar-mass black holes and confirmed general relativity in the strong-field, highly relativistic regime. For their leadership in LIGO and the observation of GW150914, Rainer Weiss, Kip Thorne, and Barry Barish were awarded the 2017 Nobel Prize in Physics.

A second watershed occurred on August 17, 2017, during the second observing run (O2), when LIGO and Virgo jointly detected GW170817, a 100-second chirp signal produced by the inspiral and merger of two neutron stars with masses between 1.17 and 1.60 solar masses. Exactly 1.7 seconds after the coalescence, the Fermi and INTEGRAL space telescopes independently detected a short gamma-ray burst, GRB 170817A. Triangulation by the three-detector gravitational-wave network localized the source to a 31-square-degree region of the southern sky, enabling optical telescopes—led by the Swope Telescope in Chile—to pinpoint the electromagnetic counterpart, AT 2017gfo, in the galaxy NGC 4993 roughly 130 million light-years (40 megaparsecs) away.³⁷ Over the following weeks, more than 70 observatories tracked the remnant across ultraviolet, optical, infrared, X-ray, and radio wavelengths.

The multi-messenger observation of GW170817 resolved several long-standing problems in astrophysics and fundamental physics. The optical and infrared light curves confirmed the existence of a kilonova powered by the radioactive decay of heavy elements synthesized via rapid neutron capture ($r$-process nucleosynthesis), demonstrating that neutron star mergers are a primary cosmic forge for elements heavier than iron, such as gold and platinum. Furthermore, the near-simultaneous arrival of the gravitational wave and the gamma-ray photons after a 130-million-year journey proved that the speed of gravity equals the speed of light to within roughly one part in $10^{15}$, ruling out numerous modified gravity and dark energy theories. By combining the luminosity distance derived directly from the gravitational-wave amplitude ("standard siren") with the electromagnetic redshift of NGC 4993, physicists also established an independent measurement of the Hubble constant, free from the calibration uncertainties of the cosmic distance ladder.

In June 2023, the NANOGrav, EPTA/InPTA, PPTA, and CPTA collaborations simultaneously published analyses of up to 25 years of millisecond pulsar observations, reporting the first compelling evidence—at statistical significances between 3 and 4.6 sigma—for a nanohertz stochastic gravitational-wave background. Exhibiting the spatial correlations of the Hellings–Downs curve, this low-frequency background is widely interpreted as the collective hum of hundreds of thousands of supermassive black hole binaries spiraling together at the centers of merging galaxies, though contributions from early-universe cosmological processes remain under investigation.

The LIGO–Virgo–KAGRA (LVK) Collaboration conducted its fourth observing run (O4) from May 24, 2023, to November 18, 2025, divided into three segments (O4a, O4b, and O4c). Enhanced by frequency-dependent quantum squeezing and optical upgrades, the network achieved unprecedented sensitivity, detecting a compact binary coalescence roughly every two to three days. Following the January 1, 2024 Noto Peninsula earthquake in Japan, which damaged KAGRA's mirror suspension systems, Japanese researchers repaired and upgraded the underground detector, returning it to joint observation for the final O4c segment in June 2025. In May 2026, the collaboration released the Gravitational Wave Transient Catalogue-5.0 (GWTC-5), incorporating 161 events from O4b and bringing the cumulative number of confirmed detections since 2015 to 390. Notable signals in GWTC-5 included GW240615, localized to an unprecedented 6 square degrees on the sky, and GW250114, a binary black hole merger recorded on January 14, 2025, with a record signal-to-noise ratio of 76.9, allowing physicists to measure three distinct vibrational ringdown modes of a remnant black hole and identify populations of second-generation black holes formed in prior mergers. Following O4, the detectors entered an upgrade phase ahead of a six-month observing run (IR1) scheduled for January 2027 and the fifth major observing run (O5) planned for the late 2020s.

To expand the terrestrial network's baseline and localization accuracy, construction is underway on LIGO-India, a 4-kilometer interferometer in the Hingoli district of Maharashtra, developed by Indian research institutions in partnership with the LIGO Laboratory.³⁸ Looking toward the 2030s and 2040s, physicists are developing third-generation ground-based observatories designed to achieve a tenfold improvement in strain sensitivity and extend observations down to 1 to 10 hertz. In Europe, the Einstein Telescope (ET) is designed as an underground, cryogenic facility arranged in an equilateral triangle of 10-kilometer arms (or twin L-shaped detectors), while the United States is developing Cosmic Explorer (CE), a pair of surface-level L-shaped interferometers featuring 40-kilometer and 20-kilometer arms. Together, these observatories will be capable of detecting virtually every stellar-mass black hole merger in the observable universe back to the cosmic dark ages prior to the formation of the first stars.³⁹

To open the millihertz band (0.1 millihertz to 1 hertz), which is inaccessible from Earth's surface, the European Space Agency (ESA), in partnership with NASA, formally adopted the Laser Interferometer Space Antenna (LISA) mission on January 25, 2024. Following the award of the prime industrial contract to OHB System AG in June 2025 and propulsion and telescope contracts to Thales Alenia Space in 2026, LISA is under hardware construction for a planned launch around 2035 aboard an Ariane 6 rocket.⁴⁰ Building on the technological success of the LISA Pathfinder demonstrator mission (2015–2017), LISA will consist of three identical spacecraft flying in a heliocentric orbit in an equilateral triangle with 2.5-million-kilometer arms, using laser interferometry to track the distances between internal free-floating gold-platinum test masses to picometer precision.⁴¹ Alongside proposed Chinese space-based interferometers such as Taiji and TianQin, LISA will observe the mergers of supermassive black holes across cosmic time, extreme mass-ratio inspirals of compact objects into galactic nuclei, and thousands of compact binary stars within the Milky Way.